Smart Pill Bottle Adherence Monitor

By /

Smart Pill Bottle Adherence Monitor

ISEF Category: Biomedical Engineering

Ready to Turn This Idea Into a Real Project?

This guide was put together with the help of AI research tools to give you a solid starting point.But a competitive science fair project lives in the details: refining your research question, fine-tuning your variables, analyzing your data, and presenting your findings like a seasoned scientist.

For next steps tailored to your interests, skill level, and timeline, work one-on-one with a MehtA+ mentor. Learn more about MehtA+ Science & Engineering Research Mentorship →

Subcategory: Biomedical Devices  ·  Difficulty: Advanced  ·  Setup: Home Setup  ·  Time: 1 to 2 Months

The Hook

Half of patients skip or misuse their medications. A pill bottle that weighs its contents and snaps a photo each time it opens can catch missed and wrong-pill events before harm happens. Add a small CNN trained on your own medication-photo dataset and the bottle becomes a smart auditor.

What Is It?

A load cell under the bottle reads the current pill weight. A cap-rotation encoder logs every open. A small camera in the cap photographs the dispensed pill.

A CNN trained on common pill shapes and colors classifies dispensed pills against the prescription. The system logs adherence events to a phone or local file.

The goal is a research auditing tool. Real medical use needs many more controls and regulatory approval.

Why This Is a Good Topic

Medication adherence is a top driver of avoidable hospital visits. Hardware is cheap and the ML problem is well-scoped. You will learn sensor calibration, embedded vision, and privacy-aware logging.

Research Questions

  • How does pill count change classifier accuracy?
  • What is the effect of lighting variability on misclassification?
  • Does load-cell trend match camera-detected dispense events?
  • To what extent does cap-rotation encoder reduce false events?
  • Which CNN architecture fits on-device with target latency?
  • How does pill-photo angle affect detection?
  • What is the effect of bottle vibration on weight readings?

Basic Materials

  • HX711 load cell.
  • OV2640 camera module.
  • ESP32 with built-in camera (e.g., ESP32-CAM).
  • Rotary encoder.
  • 3D-printed bottle housing.
  • LiPo battery.
  • Pill placebos for testing.

Advanced Materials

  • Industry-grade load cell.
  • Studio lighting rig.
  • Clinical pharmacist mentor.
  • Larger curated medication photo dataset.

Software & Tools

  • TensorFlow Lite Micro: Deploys CNN on-device.
  • PyTorch: Trains the model.
  • Arduino IDE: Programs the ESP32.
  • Python (NumPy, OpenCV): Curates the medication photo dataset.

Experiment Steps

  1. Lock the load cell and camera housing geometry.
  2. Calibrate the load cell with known weights.
  3. Build a labeled medication photo dataset.
  4. Train CNN with cross-validation.
  5. Test adherence simulation with planned errors.
  6. Report event-level precision and recall.

Common Pitfalls

  • Letting bottle vibration register as weight changes.
  • Training the CNN on only one lighting condition.
  • Ignoring partial-pill dispensing events.
  • Skipping rotation-encoder events and relying on camera alone.
  • Reporting accuracy without confidence intervals.

What Makes This Competitive

Test against at least ten realistic medications, run a within-subject simulation with planned mistakes, and report both false-positive and false-negative rates. Calibrate the load cell, fix lighting, and document the privacy story (on-device only). Add a fairness slice across pill colors.

Project Variations

  • Add a phone-app push notification for missed doses.
  • Use mmWave radar to detect bottle handling without opening.
  • Compare in-cap vs. external camera placements.

Learn More

  • PubMed: Search medication adherence smart bottle reviews.
  • NIH PubMed Central: Open-access adherence intervention papers.
  • TensorFlow Lite Micro guides: Free embedded ML tutorials.
  • NIST Image Quality Lab: Reference standards.
  • MIT OpenCourseWare: Course 6.S191 Introduction to Deep Learning.
Shopping Cart